Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation

Joanne Truong, Max Rudolph, Naoki Harrison Yokoyama, Sonia Chernova, Dhruv Batra, Akshara Rai
Proceedings of The 6th Conference on Robot Learning, PMLR 205:859-870, 2023.

Abstract

If we want to train robots in simulation before deploying them in reality, it seems natural and almost self-evident to presume that reducing the sim2real gap involves creating simulators of increasing fidelity (since reality is what it is). We challenge this assumption and present a contrary hypothesis – sim2real transfer of robots may be improved with lower (not higher) fidelity simulation. We conduct a systematic large-scale evaluation of this hypothesis on the problem of visual navigation – in the real world, and on 2 different simulators (Habitat and iGibson) using 3 different robots (A1, AlienGo, Spot). Our results show that, contrary to expectation, adding fidelity does not help with learning; performance is poor due to slow simulation speed (preventing large-scale learning) and overfitting to inaccuracies in simulation physics. Instead, building simple models of the robot motion using real-world data can improve learning and generalization.

Cite this Paper

BibTeX
@InProceedings{pmlr-v205-truong23a, title = {Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation}, author = {Truong, Joanne and Rudolph, Max and Yokoyama, Naoki Harrison and Chernova, Sonia and Batra, Dhruv and Rai, Akshara}, booktitle = {Proceedings of The 6th Conference on Robot Learning}, pages = {859--870}, year = {2023}, editor = {Liu, Karen and Kulic, Dana and Ichnowski, Jeff}, volume = {205}, series = {Proceedings of Machine Learning Research}, month = {14--18 Dec}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v205/truong23a/truong23a.pdf}, url = {https://proceedings.mlr.press/v205/truong23a.html}, abstract = {If we want to train robots in simulation before deploying them in reality, it seems natural and almost self-evident to presume that reducing the sim2real gap involves creating simulators of increasing fidelity (since reality is what it is). We challenge this assumption and present a contrary hypothesis – sim2real transfer of robots may be improved with lower (not higher) fidelity simulation. We conduct a systematic large-scale evaluation of this hypothesis on the problem of visual navigation – in the real world, and on 2 different simulators (Habitat and iGibson) using 3 different robots (A1, AlienGo, Spot). Our results show that, contrary to expectation, adding fidelity does not help with learning; performance is poor due to slow simulation speed (preventing large-scale learning) and overfitting to inaccuracies in simulation physics. Instead, building simple models of the robot motion using real-world data can improve learning and generalization.} }
Endnote
%0 Conference Paper %T Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation %A Joanne Truong %A Max Rudolph %A Naoki Harrison Yokoyama %A Sonia Chernova %A Dhruv Batra %A Akshara Rai %B Proceedings of The 6th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2023 %E Karen Liu %E Dana Kulic %E Jeff Ichnowski %F pmlr-v205-truong23a %I PMLR %P 859--870 %U https://proceedings.mlr.press/v205/truong23a.html %V 205 %X If we want to train robots in simulation before deploying them in reality, it seems natural and almost self-evident to presume that reducing the sim2real gap involves creating simulators of increasing fidelity (since reality is what it is). We challenge this assumption and present a contrary hypothesis – sim2real transfer of robots may be improved with lower (not higher) fidelity simulation. We conduct a systematic large-scale evaluation of this hypothesis on the problem of visual navigation – in the real world, and on 2 different simulators (Habitat and iGibson) using 3 different robots (A1, AlienGo, Spot). Our results show that, contrary to expectation, adding fidelity does not help with learning; performance is poor due to slow simulation speed (preventing large-scale learning) and overfitting to inaccuracies in simulation physics. Instead, building simple models of the robot motion using real-world data can improve learning and generalization.
APA
Truong, J., Rudolph, M., Yokoyama, N.H., Chernova, S., Batra, D. & Rai, A.. (2023). Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation. Proceedings of The 6th Conference on Robot Learning, in Proceedings of Machine Learning Research 205:859-870 Available from https://proceedings.mlr.press/v205/truong23a.html.

Related Material